Civil Engineering Reference
In-Depth Information
Proof.
Suppose the conditions on
a
and
b
are satisfied. We first show that for every
pair of functionals
(f, g)
∈
X
×
M
, there is exactly one solution
(u, λ)
of the
saddle point problem satisfying
+
β
−
1
1
g
,
C
α
u
≤
α
−
1
f
+
(
4
.
12
)
β
−
1
1
β
−
1
1
C
β
C
α
C
α
λ
≤
+
f
+
+
g
.
V(g)
is not empty for
g
∈
M
. Indeed, by Lemma 4.2(ii), there exists
u
0
∈
V
⊥
with
Bu
0
=
g.
u
0
≤
β
−
1
Moreover,
g
.
With
w
:
=
u
−
u
0
, (4.4) is equivalent to
a(w, v)
+
b(v, λ)
=
f, v
−
a(u
0
,v)
for all
v
∈
X,
(
4
.
13
)
b(w, µ)
=
0
for all
µ
∈
M.
By the
V
-ellipticity of
a
, the function
1
2
a(v, v)
−
f, v
+
a(u
0
,v)
attains its minimum for some
w
∈
V
with
w
≤
α
−
1
(
f
+
C
u
0
)
≤
α
−
1
(
f
+
Cβ
−
1
g
).
In particular, the Characterization Theorem II.2.2 implies
a(w, v)
=
f, v
−
a(u
0
,v)
for all
v
∈
V.
(
4
.
14
)
The equations (4.13) will be satisfied if we can find
λ
∈
M
such that
b(v, λ)
=
f, v
−
a(u
0
+
w, v)
for all
v
∈
X.
The right-hand side defines a functional in
X
, which in view of (4.14) lies in
V
0
.
By Lemma 4.2(iii), this functional can be represented as
B
λ
with
λ
∈
M
, and
λ
≤
β
−
1
(
f
+
C
u
).
This establishes the solvability. The inequalities (4.12) follow from the bounds on
u
0
,
w
, and
λ
and the triangle inequality
u
≤
u
0
+
w
.
Search WWH ::
Custom Search